Translations should be admired, not trusted.
Literature and translation are consubstantial, as Jorge Luis Borges says—and Borges as you know is always right. Their mysterious kinship is founded upon an infinity of inter- and intralingual movements and mutations: of letter into spirit (and vice versa!) and letter into letter; of the literal into the figurative; ultimately, of the literal into the literary. These transmutations break the bonds of words and things and, sometimes, the bounds of reason. In a monolingual universe, i.e., one that doesn't know or will not acknowledge the ubiquity of translation, those bonds are sacred. Violating and vitalizing the continuity of spirit, translation is sacer: at once unholy (or accursed) and—perhaps messianically—holy.
Despite recent inroads of theory into translation, translators are by nature diehard empiricists, so our work starts with what Fritz Senn calls an "inductive scrutiny" of words—what they mean, what they are made of, and what in turn they make, as they combine with their neighbors into syntactical units larger than the word, or send shoots of association, forming intricate networks. The two key words in my title are almost identical in shape, though seldom in meaning. Letters are the particulate material and the fundamental objects on the page: literature is literally made of letters. This is the first-intuitive and conjoint—definition of literature and the literary: that which is made of letters. "Avec ses vingt-quatre signes, cette Littérature exactement dénommée les Lettres, ainsi que par de multiples fusions en la figure de phrases puis le vers, système agencé comme spirituel zodiaque, implique sa doctrine propre, abstraite, ésotérique comme théologie,"  remarks Mallarmé in "La Littérature. Doctrine." He also writes of the "miracle, in the highest sense of the word" of
words led back to their origin, which is the twenty-six letters of the alphabet, so gifted with infinity that they will finally consecrate Language. Everything is caught up in their endless variations and then arises out of them in the form of the Principle. . . The book [is] a total expansion of the letter. "The Book: A Spiritual Instrument"1
The preceding two quotes contain a numerical scandal: twenty-four (letters) translated as twenty-six. English and French share but do not have the same alphabet: this asymmetry is invisible but does become a problem in the singular case of translating a text which refers to its own alphabetical element, when letters become literary. The letters of the alphabet then expose a fundamental rift: no matter how accurate or "literal" the translation is, the very English of the second quote essentially falsifies Mallarmé's doctrine propre. In other words, letters can be said to have—indeed, to be—a meaning of sorts, and this literal meaning confounds translation when it becomes literary. It takes translation to reveal this oddity. (For more, visit Borges's "Library of Babel.")
The affinity between the literal and the literary is vexed by further inconstancies of meaning: "literal" as the core of the adverb "literally" can mean its own opposite; "literally" is, in fact, often used erroneously to mean "figuratively" as, for example, when Joyce knowingly opens "The Dead" with this solecism: "Lily, the caretaker's daughter, was literally run off her feet." Notice that what is an erroneous, if common, misuse of "literally" is, in this case (i.e., in a work of art), precisely the hallmark of the literary. (The error has a characterological value: for the central intelligence in the story, Gabriel Conroy, who is something of a snob and a pedant, common and erroneous are almost synonymous.) This sliding of the literal into the figurative is crucial for a proper understanding—and translation—of any metaphor. "Literal" in "literal translation" is literal only metaphorically, not literally, and goes back to Horace's nec verbum verbo, i.e., not word for word, which has to do with syntactic rules and license, not with phonographics.
One would imagine that translators have an in-trained intuitive grasp of "literal translation"—or do they? The first commandment of our craft is Thou shalt not translate literally, precisely because literal translation is assumed to lose or destroy the literary (effect); literal translation makes the literary ludicrous. However, the same commandment against literal translation can be reversed precisely in the name of, or for the sake of, the literary, precisely according to the theory which proposes that there is some such thing as the literary. ("That focus upon expression, upon the verbal mass itself, which I have called the only essential characteristic of poetry, is directed not only to the form of the phrase, but also to the form of the word itself" reads Jakobson's definition of the literary.) In recent years what could loosely and preliminarily be called "literal" translation has risen out of disfavor and into theoretical and even practical vogue. In part, this is a cyclical turn, and has to do with the lure of the foreign (over the fish of the familiar). At its best it takes the form of comic workshop exercises in phonetic "translation" from the mere appearance of a language unknown to the translator; at its worst, it is the product of unquestioned theory used to justify questionable translation. In serious translation enterprises, the pursuit of difference through strange syntax results in a misleading sense of (modernist) experimentation where there may be none, i.e., the use of Latin or German word order in English. From the standpoint of ordinary communication, there is nothing un- or defamiliar about that order in Latin or German, of course, but things do change when you look at, say, the obligatory position of the verb at the end of the sentence in certain German constructions: for the poet and the poetologist this is a non-trivial linguistic fact. Poets, who are literally tactile syntacticians, have been known to prefer at times certain species of literal translation for the sheer feel of it.
In short, the crossings of the literal and the literary are rather peculiar, paradoxical, and elusive. My mother (herself a translator) taught me with a rap across the knuckles to beware literal translation, but corporeal punishment notwithstanding, I remain drawn to the meaning of the literal for the literary. Translating James Joyce, among others, made me keenly aware that the pursuit of the literary through the literal can range from the plausible to the preposterous; it also honed my awareness of translation's value as a double-edged tool for analysis. Nothing takes us faster to the heart of matters linguistic and metaphysical than translation: that's precisely the reason why the early Church fathers report fistfights over issues of translation; why Luther premised the Reformation upon his sense of the spirit vs the letter of scripture-in-translation or why the mullahs sentenced Salman Rushdie to death (and did, in fact, kill his Japanese translator). Translation is far more than an analytical tool: it is the medium, the mode of being, if you will, of literature.
Poetry has been defined as that which is lost in translation (conversely, translation is that which loses poetry). The adage is attributed to Frost but it predates him by a long stretch: it antedates the Romantic view of poetic language it evokes, i.e., an indissoluble, organic, indeed sacred and mysterious bond between the matter (or letter) of a given language and the spirit of its poetry. Ultimately, the spirit of this adage goes back to the Church fathers (read St. Jerome, patron of translators), and the notion of a "sacred text."
The post-Romantic/modernist view which aestheticizes (and re-sacralizes) the sacred comes down to the following crux: If the literary is coterminous with an intentional focus on language (Mallarmé, Benjamin, Jakobson et al), then the literary and the literal are one and the same. If the "literary" is identified with the spirit (or any spirit-oriented term: vision, intent, moral), then the literary is more or less translatable. If, on the other hand, the literary is indissolubly bound up with the letter (matter), then the literary is more or less untranslatable. (A strange corollary of this state of affairs is that not literary but literal translation is the impossible one, the greater illusion, the subtler fraud.)
Morgenstern's "Nightsong (or Night Hymn) of the Fish" (or perhaps "Piscine Serenade") is itself already a translation of sorts, as well as a critique of poetic reason. Dubbed "the deepest German poem" by Morgenstern himself, the poem is a parody (if that's the right word) of Goethe's "Ein Gleiches," subtitled "Wanderers Nachtlied." (The Goethe-link is implicated in the title and to a lesser extent in the prosody: Goethe's poem has a somewhat similar but irregular syllabic scheme.)
Morgenstern's poem is minimalist with a vengeance: its material is the notion of poetry as numbers. It is an almost perfect/pure representation of schematicism (and a schema of representation), a dance of steps/numbers. The poem's alphabet are the prosodic signs for long and short syllables: conventions of poetry & poetry of convention. Morgenstern poem teeters on the mini-most verge of materiality in that it is also, as a fishsong, imperceptibly audible and, as a nightsong, imperceptibly visible. If you read it as a schematization of long/short syllables, then some of its lines are unsayable/impossible. (E. g., that line of 4 short syllables?!)
Morgenstern's bride of quietness is a ravished vertical fish: sort of dead, or about to die, being caught/pulled out of its element, likely singing its last. It is more subtly dead than the supine/prone quadruped fishes of pro-/anti-Darwin bumper stickers, for if ever there was a poem that answered the call that it should "not mean/ but be" this is it.
You may think there is nothing to translate here (apart from the title which we can ignore)–and you'd be right. Except that there are several different translations of the body of the poem. What interests me is the case of Max Knight vs W. D. Snodgrass.
Snodgrass translates it by leaving it as it is, either because he believes that there is nothing to translate (which is reasonable) or because he reads it referentially, i.e., as referring to the prosodic sign for short/unaccented syllable. In light of my earlier remarks, Snodgrass's would be a sort of literal translation, pointing to the putative "facts" behind the poem. Its meaning is its pattern.
Max Knight, on the other hand, performs what one might call a literary (and self-reflexive) translation: literary in that by inverting the poem he responds to its wit with a par excellence literary gesture of his own; self-reflexive in that he bares the device of his translation qua translation. Notice that his inversion of half/!/ of the moony/scaly units of the poem runs the risk of destroying its putative reference (to long/short syllables), and perhaps changes its tonality: in the original, the breves may suggest a school of happy scales or mouths; in Knight's version the inverted breves recall the icon of tragedy. But whereas the charm of Morgenstern's poem depends on its iconic facetiousness, a neutral, aniconic reading of its signs as sheer graphic signs against the undifferentiated white of the page takes us into deeper theoretical waters: The "literal" meaning of writing, according to Jacques Derrida is "metaphoricity itself." Metaphoricity (or the use of figurative language) is, of course, a fairly common way of defining the literary. Is Morgenstern's fish, then, literature qua literature, literally?
Hermeneutics does not engage the literal materiality of the letter. In fact, what it strives for is precisely to overcome and cancel it. And so does reading. A true reader is expected to take in at a glance whole words and even combinations of words. And translation is usually based on a hermeneutic foundation. But I'm especially interested, as you've seen, in cases which compel us to re-examine our assumptions; specifically, cases where certain letter-effects—or even diacritical effects—are not accidental or ornamental but fundamental—so much so that we have to deal with a coincidence of the literal and the literary.
And even though there are no letters in the body of Morgenstern's poem, Knight's version, literally embodies the possibility and impossibility of translation. This impossible, self-contradictory condition, this paradox, is just another name for the literary. Notice that the self-reflexive translatorial gesture is made possible (or visible) by the offsetting bilingual pages. I'm no fan of bilingual editions in principle (because they tend to encourage piecemeal readings which inevitably miss the larger aesthetic effect) but in this case the bilingual edition becomes a condition of possibility, in that it exemplifies the very being of more than one fishy language.
And now I see it's time I let all of us off the hook, and lapsed back into a decent silence to become once more (as Yeats once put it) "dumber than a fish." Since nobody's last words have been able to drown out Kafka's, I'll close now, once and for all, with a particularly telling snippet from his "Silence of the Sirens"—suggesting intimate and ultimate extensions of my literary trail. (The fact that little narratorial asides like the word "namely" and the phrase "if one may so express it" survive in translation is not incidentally to be celebrated, for the passage tells us something about the fate of the telling itself.) And here it is:
The Sirens have a still more fatal weapon than their song, namely their silence
[...] and when Ulysses approached them, the potent songstresses actually did not sing
[...] But Ulysses, if one may so express it, did not hear their silence...
[Editor‘s note: The journal considered Nikolai Popov‘s paper to be quite a catch, and only regrets that it had to be scaled down to a scroll-friendlier format. We are, however, pleased to note that the original, written for the Lost and Found in Translation conference held at the University of Iowa last fall, will appear in The Iowa Review later this year.]